Embodied algorithms in vocal performance: Kinesthesis 2.0 and the transformation of player–instrument–algorithm assemblages

Abstract

This article examines Kinesthesis 2.0, an interactive system that transforms gesture and facial expression into real-time vocal processing through algorithmic integration. Drawing on over two decades of compositional practice for theatre, dance and film, we analyse how machine learning, gesture recognition and audio processing configure player–instrument–algorithm assemblages. The analysis is grounded in vocal embodiment rather than in the voice as a discrete signal: vocal sound emerges through knowledge cultivated in sustained practice, which the system extends through gestural control and algorithmic mediation. Case studies from three theatrical productions, including human–AI co-improvisation with Somax2, examine how algorithmic systems reconfigure performer agency and embodied experience while revealing productive tensions between human expressivity and computational interpretation. The methodology is qualitative and practice-based, combining technical analysis, structured performer observation and artistic reflection; observations derive from sustained artistic practice rather than controlled experimentation. The research contributes to ongoing discussions in electroacoustic and embodied performance practice by proposing that computational mediation becomes experientially seamless not through technical simplification but through cultural depth: rooting algorithmic gesture vocabularies in embedded cultural tradition allows performers to internalise the system as an extension of practice rather than as an external tool.